Magnetic vector potential
Construct B as the curl of a vector potential and distinguish gauge freedom from physical magnetic flux. Under a declared stationary localized-current/Coulomb-gauge model, derive the vector Poisson equation and its Green-kernel solution, recover Biot–Savart by taking a curl, compare symmetric and Landau gauges for uniform B, and explain why infinite-wire potentials require a finite reference radius.
- Use ∇·B=0 to justify a local B=∇×A representation and state the global/topological caveat.
- Prove B is invariant under A→A+∇χ for smooth single-valued χ.
- Apply ∇×(∇×A)=∇(∇·A)−∇²A in the magnetostatic vacuum curl equation.
- State Coulomb gauge ∇·A=0 and derive ∇²A=−μ₀J for localized steady currents.
- Use the free-space Green kernel 1/(4πR) to obtain A(r)=μ₀/(4π)∫J(r′)/R d³r′ under suitable decay.
- Take the observation-coordinate curl to recover B as the Biot–Savart integral with correct cross-product sign.
- Show ∮A·dℓ=∫B·dS for a simply supported oriented surface and explain gauge invariance of the flux.
- Compare two explicit potentials of the same uniform B using a calculable gauge function.
- Derive A_z(s)=−μ₀I/(2π)ln(s/s₀) for an infinite filament and explain the reference/gauge issue.
- Solve six exercises checking curl, divergence, gauge, units, flux and boundary limitations.
Magnetic field lines have no source or sink: ∇·B=0. A vector potential represents B by a curl rather than a gradient. The potential is not unique; its changes under a regular gauge transformation must be distinguished from measurable B and flux. We fix stationary localized vacuum currents and clear boundary assumptions before solving for A.
0104.1 — Why a vector potential exists locally
The magnetostatic magnetic Gauss law is ∇·B=0. On a simply connected regular region (subject to appropriate global topology and boundary conditions), a divergence-free field can be represented locally as the curl of some vector field A. This follows from the vector-calculus solenoidal-field construction; it is not a claim that every global topology admits the same smooth single-valued potential.
Unlike the electrostatic E=−∇V relation, a general B is not a gradient field: the stationary current law allows ∇×B=μ₀J.
0204.2 — Gauge freedom is not a new magnetic field
Let χ(r) be a sufficiently smooth, single-valued scalar on the considered region. Replace A by A′=A+∇χ. Since the curl of a gradient vanishes, this produces exactly the same B. The value of an individual component of A is therefore gauge dependent.
An open-path integral of A can change under a gauge transformation. For a closed path, ∮∇χ·dℓ=0 when χ is single-valued along the loop, so its A circulation is gauge invariant. Singular gauge functions and multiply connected domains require extra care and cannot be dismissed by a casual global argument.
0304.3 — Choose Coulomb gauge and derive vector Poisson
Use the stationary vacuum source equation ∇×B=μ₀J and substitute B=∇×A. The exact vector identity for a double curl gives ∇×(∇×A)=∇(∇·A)−∇²A. Gauge freedom allows the Coulomb-gauge condition ∇·A=0 for an appropriate localized steady-current solution.
The resulting equation is a componentwise Poisson equation. It is a differential boundary-value problem: specifying only the current without an appropriate decay or boundary prescription does not uniquely specify A, even after a partial gauge choice.
0404.4 — Free-space Green solution for localized steady currents
For a current distribution J(r′) localized in a finite region, with sufficient decay at infinity and no imposed harmonic homogeneous solution, use the free-space Green kernel satisfying ∇²[1/(4πR)]=−δ³(r−r′). Apply it to each Cartesian component of the vector Poisson equation.
This gives A as a VECTOR source integral with the same single inverse-distance kernel for every component. For a divergence-free, localized J, integration by parts makes ∇·A=0 when the surface term at infinity vanishes. The formula is not directly an infinity-normalized potential for an ideal infinite line current.
0504.5 — Recover Biot–Savart by taking the OBSERVATION curl
Take ∇ with respect to r, not the primed source coordinate r′. Since J(r′) is independent of r, ∇×[J(r′)/R]=∇(1/R)×J(r′). Because ∇(1/R)=−(r−r′)/R³, antisymmetry of the cross product converts the expression into J×(r−r′)/R³.
The sign is fixed before integration, recovering the same Biot–Savart law as EM-02.02. This is an equivalence within the stationary localized-current model, not a derivation of the full retarded fields of arbitrary time-dependent sources.
0604.6 — Stokes connects A circulation to physical magnetic flux
For an oriented smooth surface S with closed boundary C, Stokes' theorem gives ∮_C A·dℓ=∫_S(∇×A)·n̂dS=∫_S B·n̂dS. The flux is a physically defined quantity; the regular gauge transformation leaves this closed-loop integral unchanged.
For a uniform B=+B ẑ and a circular disk of radius a with its positive normal +ẑ, its flux is Bπa². This does not assert that A itself is unique or directly measurable in the same manner as B.
0704.7 — Two different potentials for the SAME uniform B
Let B=B ẑ be constant. The symmetric-gauge choice is A_sym=(−By/2,Bx/2,0); direct Cartesian differentiation gives (∇×A_sym)_z=∂x(By/2)−∂y(−By/2)=B. A second (Landau) gauge is A_L=(0,Bx,0), which also has curl +B ẑ.
Their difference is ∇χ with χ=Bxy/2. For B=0.200 T, x=0.0300 m, y=0.0400 m, the two A vectors at that single point differ numerically while describing identical magnetic force and flux.
0804.8 — Infinite wire: finite reference radius and the logarithm
The localized-source Green integral may not converge for an infinite straight filament; instead use its already established field Bφ(s)=μ₀I/(2πs). Choose a purely axial potential A=A_z(s) ẑ. In cylindrical coordinates the azimuthal curl component is (∇×A)_φ=−dA_z/ds.
Integrating with the finite reference A_z(s₀)=0 produces A_z(s)=−μ₀I/(2π)ln(s/s₀) for s>0. An infinity-zero reference is inappropriate because the logarithm diverges as s→∞. A_z's additive constant is gauge-dependent; Bφ is independent of s₀.
0904.9 — Work protocol, numerical flux, and limits
Begin by declaring stationary or time-varying sources, vacuum/material model, gauge and boundary conditions. Check whether a candidate A has the desired curl and whether the gauge difference is a true gradient. For a localized steady current, choose a free-space Green solution only when the source integral and its boundary conditions are controlled.
For a circular disk of radius a=0.0500 m in uniform B=0.200 T, Φ_B=Bπa²≈1.57×10⁻³ Wb. The closed-loop integral of either A_sym or A_L equals this flux despite different pointwise potentials. In time-dependent electromagnetic theory, E=−∇V−∂A/∂t and gauge transformations involve V as well; that full relation belongs in EM-03.
04.10 — Six exercises with worked solutions
- Compute the curl of A=(−By/2,Bx/2,0) for constant B and confirm it yields B ẑ.
- Show explicitly that A_L=(0,Bx,0) and A_sym differ by ∇(Bxy/2), and calculate both at B=0.200 T, x=0.0300 m, y=0.0400 m.
- Use the vector identity and Coulomb gauge to derive ∇²A=−μ₀J from stationary ∇×B=μ₀J.
- Take the observation-coordinate curl of A(r)=μ₀/(4π)∫J(r′)/|r−r′|d³r′ and recover the signed Biot–Savart kernel.
- For B=0.200 T through a +z-oriented circular disk a=0.0500 m, compute ∮A·dℓ and its SI units.
- For a +z infinite filament, integrate Bφ=μ₀I/(2πs) to find an axial A_z(s) vanishing at finite s₀; explain the failure of A_z(∞)=0.
Worked solutions
1. (∇×A)_z=∂x(By/2)−∂y(−By/2)=B, while x and y curl components vanish.
2. A_L−A_sym=(By/2,Bx/2,0)=∇(Bxy/2). At the given point A_sym=(−0.00400,0.00300,0) T·m, A_L=(0,0.00600,0) T·m.
3. ∇×(∇×A)=∇(∇·A)−∇²A=μ₀J; with ∇·A=0, obtain ∇²A=−μ₀J.
4. ∇_r(1/R)=−Rvec/R³; ∇×(J/R)=−Rvec×J/R³=J×Rvec/R³, hence Biot–Savart with its positive cross-product order.
5. Φ_B=Bπa²≈1.57×10⁻³ Wb = T·m²; Stokes makes ∮A·dℓ equal this positive flux for the corresponding orientation.
6. Since Bφ=−dA_z/ds, integrate to A_z=−μ₀I/(2π)ln(s/s₀) for A_z(s₀)=0. ln(s/s₀) diverges at infinity; its reference is finite and gauge-dependent.
Chapter 05 derives inductance and magnetic energy for linear quasistatic circuits, including self versus mutual inductance, flux linkage, field-energy density, and limits in lossy/dispersive magnetic materials.